Empirically-Identified Restricted Repetitive Behavior Domains: Informing DSM-6

Friday, May 13, 2016: 5:30 PM-7:00 PM
Hall A (Baltimore Convention Center)
T. W. Frazier1 and A. Y. Hardan2, (1)Cleveland Clinic Center for Autism, Cleveland, OH, (2)Stanford University, Stanford, CA

Restricted, repetitive behavior (RRB) is one of the two symptom domains used to diagnose autism spectrum disorder. Diagnostic and Statistical Manual of Mental Disorders – Fifth Edition (DSM-5) criteria include 4 sub-domains of RRB, including repetitive motor/speech, sameness/inflexibility, restricted interests, and abnormal sensory experience. In contrast, empirical studies of RRB suggest two broad domains – insistence on sameness and repetitive sensory motor behavior. However, diagnostic criteria are based on limited empirical work and previous factor analytic studies of RRB have been restricted to single instruments, limited item sets, or single samples.


To empirically-identify RRB sub-domains using confirmatory factor analyses based on two instruments within each of two separate samples.


Parent-reported symptom data were obtained from 2,643 children with ASD included in the Simons Simplex Collection (SSC) using the Social Responsiveness Scale and Repetitive Behavior Scale-Revised and 3,907 children with ASD from the Interactive Autism Network (IAN) using the Social Responsiveness Scale and Social Communication Questionnaire. In each dataset, items were a priori chosen to represent one of five constructs: repetitive motor behavior, insistence on sameness, restricted interests, sensory interests, and sensory sensitivities. A series of confirmatory factor analyses (CFA) were conducted using item-level data with increasingly complex models, including a 1-factor model merging all items, a 2-factor model merging repetitive motor with sensory interests and the remaining three constructs, three 3-factor models iteratively separating restricted interests, sensory interests, and sensory sensitivities from the prior 2-factor model, and a 5-factor model with all five constructs as separate factors. Improved fit across increasingly complex models was evaluated using changes in the Bayesian Information Criterion (BIC), Root Mean Square Error of Approximation (RMSEA), and Confirmatory Fit Index (CFI).


In both samples, the 5-factor model fit best (smallest absolute improvement: ΔBIC=424, ΔRMSEA=.004, ΔCFI=.024). Absolute fit was marginal for the 5-factor model, as would be expected by multi-item factor analyses without cross-loadings based on complex psychopathology questions. Factor correlations indicated that all RRB constructs had positive, but modest, inter-relationships (SSC r=.06-.22; IAN r=.17-.37), with the exception of the very large relationship between repetitive motor behavior and sensory interests (SSC r=.46, 95%CI=.41-.51; IAN r=.59, 95%CI=.55-.63). Qualitative examination of item loadings revealed that the restricted interests and sensory sensitivities factors, while clearly distinct from other factors, included only a few salient loadings.

Conclusions:  Additional large sample studies are needed to confirm and extend the present findings. If confirmed, future revisions of diagnostic criteria may benefit from separating sensory interests and sensory sensitivities, as these two constructs appear only modestly correlated. The present data indicate that sensory interests may be a statistically separable from repetitive motor behavior, although the high correlations observed across samples suggests that future investigations evaluate the clinical and diagnostic utility of this distinction. Only a small number of items were available to evaluate restricted interests, yet this construct clearly separated from other insistence on sameness items, supporting current diagnostic criteria. Future instrument revisions should focus on beefing up measurement of restricted interests and sensory sensitivities.